We should think about how extractors can report their desired formats at some point (thinking e.g. if we'll have issues with pandas 3 etc) _Originally posted by @ml-evs in https://github.com/datatractor/beam/pull/3#pullrequestreview-2294132022_